SE PhD Final Defense: Ehsan Sabouni

SE PhD Final Defense: Ehsan Sabouni

TITLE: Safety-Critical Multi-Agent Control and Coordination for Autonomous Vehicles: Barrier Functions, Learning, and Deployment

ADVISOR: Christos Cassandras (ECE, SE)

COMMITTEE: Roberto Tron (ME, SE), Ioannis Paschalidis (ECE, BME, SE), Wenchao Li (ECE, SE) - Chair:

ABSTRACT: The emergence of Connected and Automated Vehicles (CAVs), coupled with advancements in traffic infrastructure, promises to address longstanding transportation challenges, including accidents, congestion, energy inefficiency, and environmental pollution. Achieving this vision depends critically on effective traffic management, particularly at bottlenecks such as intersections, roundabouts, and merging roadways. This dissertation advances safety-critical multi-agent control and coordination for autonomous vehicles through a series of integrated contributions. First, event-triggered and self-triggered control strategies are developed for CAVs using Control Barrier Functions (CBFs) to enforce hard safety constraints in traffic conflict zones. These methods address the infeasibility of solving Quadratic Programs (QPs) by adaptively updating control actions based on system events, rather than fixed time steps. Second, we consider the problem of merging an autonomous vehicle (AV) into flows of Human-Driven Vehicles (HDVs). An optimal index policy is derived to minimize travel time and energy, and the framework is extended to arbitrary CAV penetration rates through a hierarchical approach that determines safe merging sequences and performs decentralized motion planning under safety constraints. Third, a bilevel Reinforcement Learning (RL) framework is proposed to automate parameter tuning within an MPC-CBF architecture. The RL agent learns both cost weights and CBF parameters without requiring backpropagation through the optimization layer. The learned controllers are validated in a multi-agent merging scenario and deployed in a Hardware-in-the-Loop (HIL) Smart City Testbed with both physical robots and simulated agents. Finally, we introduce HMARL-CBF—a Hierarchical Multi-Agent Reinforcement Learning framework combining high-level cooperative skill selection with low-level CBF-based control. This approach enables scalable coordination and strong safety guarantees in dense multi-agent traffic scenarios. Together, these contributions establish a scalable and adaptive framework for deploying CAV technologies in complex, mixed-autonomy environments. By integrating model-based safety, learning-based adaptation, and real-world validation, this dissertation lays a foundation for safe and efficient traffic systems.